An auxin-mediated ultradian rhythm promotes root regeneration in Arabidopsis

Ultradian rhythms have been proved to be critical for diverse biological processes. However, comprehensive understanding of the short-period rhythms remains limited. Here, we discover that leaf excision triggers a gene expression rhythm with ~ 3-h periodicity, named as the excision ultradian rhythm (UR), which is regulated by the plant hormone auxin. Transcriptome analysis found more than 4,000 excision UR genes which are diverse in terms of biological function. Promoter–luciferase analyses showed that the spatiotemporal patterns of the excision UR were positively associated with de novo root regeneration (DNRR), a post-embryonic developmental process. Genetic studies showed that EXCISION ULTRADIAN RHYTHM 1 (EUR1), which encodes ENHANCER OF ABSCISIC ACID CO-RECEPTOR1 (EAR1), an abscisic acid signaling regulator, was required to generate the excision ultradian rhythm and promote root regeneration. Moreover, the ear1 mutant exhibited the absence of auxin-induced excision UR generation and partial failure to rescue DNRR. These results demonstrate that leaf excision activates EAR1-mediated excision UR and reprograms the expression of a large set of genes involved in DNRR.

plants, and this is frequently observed in nature 17 . Excised Arabidopsis leaves, which can regenerate adventitious roots (ARs) at the excision site on a hormone-free medium, are frequently used as a model system to imitate natural conditions and investigate the molecular mechanisms underlying de novo root regeneration (DNRR) 18 . DNRR is a highly complex process that involves time-evolving regulatory networks with a series of cell fate transition, division and differentiation steps which require reprogramming of large set of gene expression 19,20 . Despite recent extensive studies, the regulatory mechanisms of gene expressions underlying the DNRR process are not fully understood.
The results presented here show a new regulatory layer that a UR triggered by leaf excision promotes DNRR by resetting gene expression patterns, thereby assisting the cells at the excision sites to reorient from their predetermined differentiated cellular states toward new fates.

Leaf excision evokes an auxin-mediated UR
We previously used the re y luciferase (LUC) reporter gene 21,22 to track the promoter activities of circadian clock-regulated genes, including ORE1, in transgenic Arabidopsis. The 3 rd or 4 th rosette leaves were excised from 21-day-old plants grown under long-day conditions (16 h light/8 h dark), and LUC activity was monitored at 30 min intervals using a CCD camera under continuous white light at 22°C. (Fig.   1a). Transgenic ORE1::LUC plants showed short period rhythms (Fig. 1b). We used wavelet analysis, which is suitable for time-frequency data 23 , to determine whether periodicity resulted from an endogenous biological rhythm. As the ORE1 promoter exhibited a circadian rhythm (CR) as well as UR, we separated the circadian component by reconstructing the smoothed circadian signal from the original oscillating pattern. Consequently, the wavelet spectrum exhibited a ~24 h period CR (Fig. 1c). Subtraction of the CR wavelet from the original oscillating pattern revealed an additional UR with a ~3 h period (Fig.  1d). ORE1, CCA1, PRR7 and CAB2 promoter activities in excised leaves showed similar periods of ~3 h, with various wavelet powers ( Fig. 1e and Supplementary Fig. 1). When a threshold of 1.0 was established to discriminate the UR from noise (Fig. 1e, red line), ORE1 promoter activity showed a signi cant wavelet power. We further characterized this ~ 3 h rhythm using ORE1 promoter activity.
We used wavelet analysis of ORE1 promoter activity to determine whether the UR was present in intact leaves and other excised organs. Attached leaves did not exhibit a UR in ORE1 promoter activity ( Supplementary Fig. 2a-c,n). Next, we examined ORE1 promoter activity in 7-day-old whole seedlings and in excised shoot apices, cotyledons, hypocotyls and roots ( Supplementary Fig. 2d). The UR was detected in excised cotyledons but not in the other samples ( Supplementary Fig. 2e-i,n). We also tested organs excised from bolted plants ( Supplementary Fig. 2j). The UR was present in excised cauline leaves but not in excised owers or stems ( Supplementary Fig. 2k-n). Thus, the ORE1 UR is leaf speci c and is evoked upon its excision. The ~3 h period UR was therefore named the 'excision UR'.
To gain insight into the physiological function of the excision UR, we examined its temporal and spatial expression patterns in excised leaves. A signi cant excision UR wavelet power was observed from ~19 h after excision and maintained until ~ 60 h before dampening over time (Fig. 1f). The excision UR thus functioned as a transient response to excision. The highest level of ORE1 promoter activity with robust oscillations was observed in the petiole region close to the excision site (Fig. 1g). By contrast, ORE1 promoter activity persisted across the whole area of an attached leaf (Supplementary Movie 1). The localized and transient nature of the UR proximate to the excision site supported the conclusion it was a response to leaf excision.
Excised Arabidopsis leaves undergo a drastic developmental shift toward DNRR at the excision site. Bỹ 12 h after excision, auxin is produced in converter cells and transported to the vasculature near the wound site, where it is involved in further DNRR processes 19,20,24,25 . These studies suggested the excision UR might be related to the production of auxin as an excision response. To test this, we monitored auxin responses in vivo and in real time using DR5::LUC transgenic plants 5 . DR5::LUC expression exhibited an excision UR with a signi cant wavelet power (Fig. 1h,i). Activity was enriched in the petiole region (Fig. 1j). Furthermore, the average wavelet power of the ORE1 promoter excision UR was reduced by yucasin, an auxin biosynthesis inhibitor 26 , and rescued by exogenous auxin (Fig. 1k), indicating that auxin positively regulated the excision UR in excised leaves. However, treatment of young, excised leaves with exogenous auxin did not signi cantly affect the UR wavelet power (Fig. 1k), suggesting that endogenous auxin levels were su cient for UR generation in young leaves. Leaf excision thus generated an auxin-regulated excision UR.
Function and expression of a large set of excision UR genes  Table 2). The excision UR was therefore involved in resetting various hormonal signaling pathways. Several metabolic pathways were also enriched among the excision UR genes. The excision UR thus predominantly involved genes acting in hormone signal transduction pathways and multiple metabolic processes.
DNRR at the leaf excision site involves a complex array of regulatory genes. Auxin plays an essential role in this process 18-20,24,25 : of 35 DNRR-associated genes identi ed previously, ten, of which seven were auxin-related, showed the UR expression pattern (Supplementary Table 3). The excision UR thus regulated the expression of genes involved in auxin-related DNRR. Auxin regulates a root clock, which produces oscillations in gene expression with a ~6 h period for prebranch site production 5 . We compared the genes showing ~3 h period UR with microarray data from the root clock to determine whether these different URs shared common molecular components. Although the two datasets showed little overlap (< 7%), YUCCA 9 (YUC9) and AUXIN RESPONSIVE FACTOR 7 (ARF7) were common to both ( Supplementary  Fig. 3). Both are auxin-related genes involved in DNRR, suggesting that, although the two URs controlled distinct sets of genes, they shared part of the auxin-mediated regulatory pathways.
The effect of the excision UR on auxin-related genes was con rmed using promoter-reporter assays. LUC activity in transgenic plants expressing PIN-FORMED 3 (PIN3)::LUC, ARF7::LUC or AUXIN SIGNALING F-BOX 2 (AFB2)::LUC showed robust excision UR ( Fig. 2d-g). Reciprocal regulation between auxin and the excision UR led us to test whether the auxin signaling pathway was involved in generating the UR. We screened the oscillating auxin-related genes ARFs (ARF4, 7, 8 and 10), PIN3, AFB2 and TRANSPORT INHIBITOR RESPONSE 1 (TIR1) by examining patterns of ORE1::LUC expression in loss-of-function mutants, as well as in a gain-of-function mutant of SHORT HYPOCOTYL 2 (SHY2). Excised leaves from all mutants showed a robust excision UR ( Supplementary Fig. 4), suggesting it was either generated upstream of auxin signaling or was genetically separate from the auxin signaling pathway.

Positive correlation between excision UR and DNRR
As both the excision UR and DNRR were induced by leaf excision and controlled by auxin, we assessed the relationship between robustness of the excision UR and e ciency of DNRR in excised leaves under various conditions. DNRR is highly sensitive to the age of an excised leaf 18,28 , with aged leaves exhibiting a marked reduction in DNRR capacity. The excision UR and DNRR e ciency were examined in leaves excised from plants of different ages. The average wavelet powers of the excision UR were robust in the 4 th leaf from 17-or 21-day-old plants, but gradually decreased with leaf age; the excision UR was not detectable in leaves from 28-day-old plants ( Fig. 3a). The e ciency of DNRR was positively correlated with the trend in excision UR (Fig. 3b,c). Thus both the excision UR and DNRR were highly sensitive to leaf age, and occurrences of the two events were correlated.
DNRR is sensitive to light conditions, as excised leaves form roots under light conditions but not in the dark without sucrose 18 . We therefore examined the effect of varying light intensity on excision UR robustness and DNRR e ciency. The average wavelet powers of the excision UR were highest under photosynthetically active radiation (PAR) of 20 mmol m -2 s -1 , and were reduced in the dark and under lower or higher light intensities (Fig. 3d). Similarly, DNRR was most e cient under PAR of 20 mmol m -2 s -1 (Fig. 3e). Excised leaves did not produce any roots under darkness due to dark-induced senescence, a rapid ageing process (Fig. 3f). Leaves exposed to lower and higher light intensities remained green for 12 days after excision but showed reduced DNRR e ciency (Fig. 3e, Fig. 5a, b). Genetic analyses revealed that all four candidates were recessive mutants. M21, M23 and M83 belonged to the same complementation group, whereas M38 formed a second distinct complementation group ( Supplementary Fig. 5c). The mutations were named EXCISION ULTRADIAN RHYTHM (EUR), and the rst and second complementation groups were named EUR1 and EUR2, respectively. All four mutants exhibited delayed initiation of ARs relative to wild type ( Fig. 4f and Supplementary Fig. 5d). These results supported a causative association of UR with DNRR, as the four mutants belonged to two independent complementation groups, and yet controlled the excision UR and DNRR simultaneously.
The presence of three eur1 mutant alleles in one complementation group facilitated molecular analysis by whole-genome sequencing (WGS). The WGS data of ORE1::LUC (parental line) were compared with that of a pool of F 2 homozygous mutant progeny, which showed no excision UR, obtained by backcrossing M21 or M83 with ORE1::LUC. Only one gene, ENHANCER OF ABSCISIC ACID (ABA) CO-RECEPTOR1 (EAR1), harboured common intragenic single nucleotide polymorphisms (SNPs) in both the M21 and M83 mutants ( Supplementary Fig. 5e). The WGS results were validated by sequencing the EAR1 coding sequence in M21, M83 and M23 ( Supplementary Fig. 5f). In M21 and M23, tryptophan residues at amino acid positions 112 and 52 were changed to nonsense codons, whereas glycine-157 was changed to glutamate in M83 (Fig. 4g). The mutant alleles in M21, M23 and M83 were named eur1-11, eur1-12 and eur1-13, respectively. To con rm that EAR1 was the gene responsible for the excision UR, complementation lines (COM-9, COM-24) were generated by expressing an EAR1-GFP fusion construct under the control of its cognate promoter (EAR1::EAR1-GFP) in the eur1-11 mutant background. The expression of EAR1::EAR1-GFP rescued both the impaired excision UR and reduced DNRR e ciency phenotypes of eur1-11 ( Supplementary Fig. 6). These results indicated that EAR1 corresponded with the eur1 mutations and was a positive regulator of excision UR in excised Arabidopsis leaves.
EAR1 is a negative regulator of ABA signaling, and the core EAR1 141-287 fragment is su cient for EAR1 function in ABA responses 29 . We tested whether the core fragment of EAR1/EUR1 generated the UR in a similar manner by using an insertion line of EAR1 (SALK_108025, eur1-14), in which the T-DNA is inserted at position 1,338 of AT5G22090 (Fig. 4g), keeping the core fragment intact. Unlike the other eur1 mutant alleles, ABA responses, inhibition of germination and root growth in eur1-14 resembled those of wild-type plants (Fig. 4h,i), con rming previous reports 29 . Notably, however, both expression of the excision UR and DNRR e ciency were impaired in eur1-14 leaves (Fig. 4j,k), suggesting that EAR1/EUR1-mediated excision UR generation and AR formation are separate from canonical ABA signaling.
Auxin-induced generation of the excision UR via EAR1/EUR1 promotes DNRR As the EAR1/EUR1 controlled both the excision UR and DNRR, we investigated the link between these two phenomena. We performed time-course RNA-seq analysis of the petiole regions of wild-type and eur1-11 mutant leaves collected at 0, 24, 48, 72 and 96 h after excision. This revealed that 9,754 genes were differentially expressed between wild type and eur1-11. These differentially expressed genes (DEGs) were categorized into 12 clusters according to the similarity between their expression pro les (Supplementary Fig. 7 and Supplementary Table 4). Interestingly, the expression pro les of genes in cluster 2, which contained EAR1/EUR1, resembled the pattern of excision UR wavelet power (Fig. 5a). To gain further insight into the role of EAR1/EUR1 in DNRR, we performed GO and KEGG enrichment analyses of the 325 genes belonging to cluster 2. These genes were strongly enriched in GO/KEGG terms related to auxin and development (Fig. 5b), suggesting that EAR1/EUR1 promoted DNRR via an auxin-mediated molecular mechanism. Indeed, the DNRR-associated genes found in cluster 2 included key genes required for auxin biosynthesis and transport, and for auxin-mediated cell fate transition, such as YUC8, YUC9, PIN2 and WUSCHEL-RELATED HOMEOBOX 11 (WOX11) 19,24 (Fig. 5c). The absence of EAR1/EUR1 altered auxin signaling in the petiole region upon excision, which may have changed the expression of genes involved in cell fate determination and resulted in reduced DNRR e ciency.
DNRR occurs at the site of excision from the petiole. Excision UR expression was the strongest at the petiole, which correlated positively with DNRR. We therefore examined the spatial and temporal regulation of EAR1/EUR1 in EAR1::EAR1-GFP plants. The uorescence signal was absent in the petiole region at 0 days after excision (DAE), but was visible from 1 DAE and most abundant at 2 DAE (Fig. 5d), indicating that the changes in EAR1/EUR1 levels coincided with expression of the excision UR.
Exogenous application of auxin rescues DNRR in aged leaves 18 . As the excision UR was also regulated by auxin (Fig. 1f) and reduced in aged leaves (Fig. 3a), we hypothesized that auxin might induce the excision UR through EAR1/EUR1 and rescue DNRR e ciency in aged leaves. To test this, we applied 10 µM IAA to 4 th rosette leaves excised from aged (24-day-old) wild-type and eur1-11 mutant plants, and measured robustness of the excision UR and DNRR e ciency. Exogenous auxin treatment rescued the excision UR wavelet power in aged wild-type leaves but not in aged eur1-11 leaves (Fig. 5e), indicating that EAR1/EUR1 was required for auxin-induced excision UR generation. The DNRR e ciency of aged wild-type leaves was also fully rescued by auxin; however, aged eur1-11 mutant leaves showed lower DNRR e ciency than their wild-type counterparts (Fig. 5f), indicating that the EAR1/EUR1-mediated excision UR was necessary to promote DNRR, although auxin could induce DNRR independently. All these results suggest that leaf excision triggers an endogenous oscillation in gene expression that promotes root regeneration, and this process is regulated by reciprocal interactions between auxin and EAR1/EUR1 (Fig. 5g).

Discussion
Here, we show that, in Arabidopsis, leaf excision activates EAR1/EUR1, which acts via auxin to generate a transient excision UR that promotes AR formation. Like other ultradian rhythms in gene expression, such as the root branching rhythm in Arabidopsis 5 , segmentation and somitogenesis in Drosophila 30 , this excision UR is involved in a developmental process, DNRR. In addition, like root clock 5 and segmentation clock 31 , there might be a ultradian clock to regulate this excision UR, in which EUR1 and EUR2 play a role as core clock genes. However, the excision UR is not associated with a spatially periodic modular developmental pattern. Instead, it is evoked de novo at the petiole region of excised leaves and is observed transiently after excision. Thus, the latent and transient excision UR has a unique oscillatory feature. DNRR is a highly complex process that involves regulatory networks that change over time and show three distinct phases 19,20 . The time-frame of the excision UR overlapped with phase II (auxin accumulation) and phase III (cell fate transition) (Fig. 1f and Supplementary Table 3). Therefore, cells undergo fate transition when the EUR is robust. This timing is indicative of the role of the latent and transient excision UR in biological processes. In addition, expression of cell fate transition genes was altered in eur1-11 mutants (Fig. 5c). Rhythmic gene expression at the excision site may serve as a means of resetting and reprogramming gene expression to facilitate cell fate transition. This would resemble the situation in lateral root development, in which oscillatory behaviour of some genes is associated with cell fate transition in response to lateral root initiation 5,32 . URs with frequent information may temporally govern gene expression to precisely control cell fate transitions during development in plants.
Robustness of the excision UR was affected by developmental stage of leaves and environmental signals such as light intensity, which also in uence DNRR e ciency (Fig. 3). As plants age, gradually increased transcription factors such as SQUAMOSA PROMOTER BINDING PROTEIN-LIKE (SPL) 2/10/11 and ETHYLENE INSENSITIVE 3 (EIN3) repress root regeneration by inhibiting auxin biosynthesis and expression of cell fate transition genes, respectively 33,34 . Auxin, a major hormone in DNRR, was required for generation of the excision UR (Fig. 1k) and also rescued the excision UR in aged leaves (Fig. 5e). This indicates that the excision UR as well as DNRR are positively regulated by auxin which level is gradually decreased along with the age of leaves. Proper light intensity was required for optimal generation of the excision UR (Fig. 3d). This may be caused by an imbalance in carbohydrate concentration, which is otherwise required for optimal DNRR (Fig. 3e,f). Previous study showed that, in excised leaves, sucrose is required in the dark to regenerate ARs, but somewhat represses root regeneration in the light 18 , suggesting that an appropriate amount of carbohydrate is necessary for optimal root regeneration as an energy source. Lower robustness of the excision UR in the dark (Fig. 3d) might also be caused by depletion of energy which can be made by photosynthesis in the light. As only leaves can make enough energy source via photosynthesis in the light, leaf-speci c occurrence of the excision UR ( Supplementary   Fig. 2) supports this explanation.
Leaf excision and subsequent DNRR processes are largely integrated by the interplay of several hormones, including early signaling by the wound hormone jasmonic acid followed by various auxin, cytokinin and ethylene 35 . This is consistent with the KEGG pathway analysis of the excision UR transcriptome (Fig. 2c) as the excision UR is associated with DNRR. However, the role of ABA signaling components in DNRR has been rarely discussed to date. One of the regulators of the excision UR identi ed from genetic screening was EAR1/EUR1, previously known as a negative regulator of ABA signaling 29 . Interestingly, EAR1/EUR1 is involved in canonical ABA responses, but the excision UR mediated by EAR1/EUR1 may be generated by a different molecular mechanism (Fig. 4h-k), which is positively regulated by auxin. EAR1/EUR1 controls genes from the auxin pathway, including auxin biosynthesis genes, resulting in a reciprocal positive feedback loop between auxin and EAR1/EUR1 (Fig.   5g). ABA is generally considered as a negative regulator of AR formation 35 . Therefore, although EAR1/EUR1-mediated excision UR generation and root regeneration was decoupled from canonical ABA responses, EAR1/EUR1 may also regulate ABA signaling during DNRR by activation of the ABA coreceptor phosphatases that negatively regulate ABA signaling, and to evoke the EUR at the excision site. Consistent with these, the expression of EAR1/EUR1 is activated at the excision site of petiole, as would be expected for the petiole excision site to be competent for cell fate transition and division. Further studies to identify more components, such as other eur mutants or factors interacting with EAR1/EUR1, will improve our understanding of the regulatory mechanisms underlying the excision UR and DNRR.

RNA-seq and functional prediction
Library construction and sequencing were performed using Illumina Hiseq 2500 platform for detecting oscillation genes and using Illumina NovaSeq 6000 platform for eur1-11. Raw reads were checked quality and trimmed using FastQC 42 , and the trimmed reads were mapped to the Arabidopsis thaliana genome (TAIR10) using STAR 43 . After alignment, the gene-level raw count data les were generated using HTSeq 44 and normalized using edgeR's TMM algorithm 45 . The differential gene expression was analyzed by the multifactor generalized linear model (GLM) approach in edgeR with replicate number added as a factor to the GLM to mitigate for a batch effect. The ltered genes with a p-value under 0.05 were considered as differential expressed genes. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway were performed by g:Pro ler for computing multiple hypothesis testing corrections (g:scs < 0.05) 46 . The ReViGO was used to summarize and visualize the list of signi cantly enriched GO terms based on semantic similarities (allowed similarity: 0.5) 47 . The RNA-seq data used in this study have been deposited in the Gene Expression Omnibus (GEO: http:/www.ncbi.nlm.nih.gov/geo/) and assigned the identi er accession GSE157230 and GSE158133.

Detecting oscillation genes
The RNA-seq dataset for UR detection (synchronized RNA-seq dataset) was analyzed by bioinformatics tools (described above). The MetaCycle R package which incorporates ARSER, JTK CYCLE and Lomb-Scargle 27 was used to detect rhythmic genes from Synchronized RNA-seq data. The ultradian rhythms were detected with parameters: minper 2 h and maxper 5 h. The genes with cut-off (p-value < 0.05) based on meta2d results were de ned as UR oscillating genes.
Clustering for differential expressed genes in eur1-11 The expression values of DEGs were analyzed by the tri-cluster system, TimesVector 48 for the relationship between time series and pattern of DEGs.
ABA seed germination and primary root growth assay with Tukey's post hoc test. Data points with different letters indicate statistically signi cant differences between groups (P < 0.01).

Figure 2
Transcriptomic and functional analysis of the excision UR genes. a, Heat map showing the expression levels of genes oscillating over time. Yellow and blue indicate higher and lower relative expression, respectively. b, Gene ontology enrichment analysis of the excision UR genes. Bars represent numbers of genes and color represents the p value. c, KEGG enrichment analysis of the excision UR genes. Dot size indicates the number of genes, and dot colour represents the P-value. d,f, Analysis of ARF7, PIN3 and AFB2 promoter activities using the LUC reporter. Graphs show data from three representative samples. The graphs (upper panels) show measurements from three representative samples (n = 24) and the wavelet spectrum plots (lower panels) show merged wavelet power plots of all samples with low transparency. g, Average wavelet powers of the excision UR of ARF7, PIN3 and AFB2 (n = 24 leaves). Centre line: median; bounds of box: 25th and 75th percentiles; whiskers: 1.5 × IQR from 25th and 75th percentiles. Statistical signi cance was determined by one-way analysis of variance (ANOVA) with Tukey's post hoc test. Data points with different letters indicate statistically signi cant differences between groups (P < 0.01).